AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Memory-Efficient LLM Training with Dynamic Sparsity: From Stability to Practical Scaling

Qiao Xiao, Boqian Wu, Patrik Okanovic, Tomasz Sternal, Maurice van Keulen, Elena Mocanu, Mykola Pechenizkiy, Decebal Constantin Mocanu, Torsten Hoefler

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Neural scaling laws · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add optimizer-state warm-up to Adam-based DST (ablation)SET + Adam · C4 LLaMA-350M ablation (d=0.1)perplexity: 28.98 -> 27.67Improved
add density-aware LR scaling to Adam-based DST (ablation)SET + Adam · C4 LLaMA-350M ablation (d=0.1)perplexity: 28.98 -> 28.20Improved
add LR scaling to warm-up (full SMET) (ablation)SMET · C4 LLaMA-350M ablation (d=0.1)perplexity: 27.67 -> 25.85Improved
add warm-up to LR scaling (full SMET) (ablation)SMET · C4 LLaMA-350M ablation (d=0.5)perplexity: 24.79 -> 21.61Improved
use warm-up length W=10 (ablation)SMET · C4 LLaMA-350M density 0.2perplexityImproved
use topology update interval of 100 steps instead of 50 or 300 (ablation)SMET · C4 LLaMA-350MperplexityImproved
replace SET with SMET (random regrowth)SET · C4 sparse pre-training (d=0.1)perplexity: 28.96 -> 25.85Improved
replace RigL with SMET_grad (gradient regrowth)RigL · C4 sparse pre-training (d=0.1)perplexity: 29.90 -> 26.50Improved
dynamic sparse training with RigL versus static sparseStatic Sparse · C4 sparse pre-training (d=0.5)perplexity: 22.90 -> 25.75Hurt
SMET versus static sparseStatic Sparse · C4 sparse pre-training (d=0.5)perplexity: 22.90 -> 21.61Improved
SMET versus GaLore at 1BGaLore · C4 LLaMA-1Bperplexity: 15.64 -> 15.61No clear effect
SMET versus GaLore at 350MGaLore · C4 LLaMA-350Mperplexity: 18.95 -> 18.90Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy